alpha dose rate of 0.03 Gy kyr−1 (ref.^56 ) and 0.72 Gy kyr−1 (refs.^57 ,^58 ) were
used for the 90–125-μm quartz and feldspar samples, respectively
(owing to the radioactive decay of^40 K and^87 Rb), which were made
assuming K (12.5 ± 0.5%)^57 and^87 Rb (400 ± 100 μg g−1)^58 concentrations,
and was included in the total dose rate. Cosmic-ray dose rates were
estimated from published relationships^59 , making allowance for the
sediment overburden at the sample locality (about 0.5-4.0 m), the
altitude (around 60 m above sea level) and geomagnetic latitude and
longitude (7° and 111°) of the sampling site. The total dose rate was
calculated using a long-term water content of 5–15 ± 2%, which is close
to the measured (field) water content of 5–15%. High-resolution gamma
spectrometry of the powdered sediment samples was also conducted
to test for disequilibrium within the uranium decay chain (Supple-
mentary Table 7).
(^40) Ar/ (^39) Ar dating of a pumice lens in the Sembungan terrace. 40 Ar/ (^39) Ar
dating of single and multiple hornblende crystals was conducted on a
pumice lens (sample SS59B) taken from the middle of the Sembungan
terrace, to provide the eruption age of the pumice and a maximum age
of deposition for the terrace. The 10-cm-thick pumice lens was exposed
in the western quarry wall over a length of 2.2 m at a depth of 2.0 m below
datum, and 2.1 m above the base of the terrace fill. Euhedral hornblende
crystals up to 1 mm in length were hand-picked under a binocular mi-
croscope and loaded into wells in aluminium sample discs (diameter 18
mm) for neutron irradiation, along with the astronomically calibrated
1.185-million-year-old Alder Creek sanidine^60 as the neutron fluence
monitor. Neutron irradiation was performed for 0.25 h in the cadmium-
shielded CLICIT facility at the Oregon State University TRIGA reactor.
Argon isotopic analyses of the gas released by laser fusion of horn-
blende crystals (Supplementary Table 12) was done on a fully auto-
mated, Nu Instruments Noblesse multi-collector noble-gas mass
spectrometer, using previously documented instrumentation and
procedures^61 ,^62.
Estimating downcutting rates. We used the terrace chronology to
establish downcutting rates for the Solo Valley in the Kendeng Hills.
These rates were established by dividing the age of each alluvial ter-
race (in thousands of years) (Supplementary Table 6), by the distance
of downcutting between each terrace (mm) to provide a downcutting
rate (in millimetres per thousand years (mm kyr−1)) for each terrace
level. Then, the age of the highest terrace was divided by the total dis-
tance to the river (66 m) to estimate an overall rate of downcutting (in
mm kyr−1) for the entire valley (Supplementary Information section 15).
The mean errors associated with these red thermoluminescence and
pIR-IRSL ages have been propagated through to the final downcutting
rate, to obtain an error margin of ±8%.
Establishing the fossil context
The strategy for the Ngandong excavation consisted of locating the
backfilled edges of the original excavated area along the margins of
the original Netherlands Indies Survey reserve, using maps produced
during the 1931–1933 excavations^3. Our 115-m^2 excavation footprint
encompassed several pits that paralleled the edges of the excavation
reserve. We identified five lithofacies, comprising the terrace fill at
Ngandong, and developed a composite cross-section that illustrated
lateral facies relationships and the context of the bone bed within the
sequence of deposits (Fig. 2 , Supplementary Table 2). Attempts to
date fossil teeth from facies A and facies C using radiocarbon were
unsuccessful, as was initial optically stimulated luminescence dating
of quartz from sediments. However, red thermoluminescence and pIR-
IRSL dating of quartz and feldspars, respectively, from overbank terrace
deposits, and U-series dating of fossil teeth and bones proved effective.
Laser-ablation U-series dating of fossil bone and teeth. Fifteen
bone samples for U-series dating were collected during the 2010 field
season (sample numbers 2146, 2178, 2190, 2216, 2269, 2284, 2286, 2291,
2319, 2330, 2331, 2388, 2404, 2476 and 2481) from facies A and facies C.
Figure 2 gives the relative positions of these samples. Laser-ablation
mass spectrometry to measure U-series isotopes along the cross-
sections of these dense, mammalian long-bone fossils was carried
out at the Australian National University^39 ,^63 –^66 (Extended Data Fig. 9).
Uranium and thorium concentrations were derived from repeated
measurements of the NBS-610 standard, uranium-isotope ratios from
the dentine of a rhinoceros tooth from Hexian^67. Spot analyses were
used, which have the advantages over continuous tracks by being able to
avoid pores and optimize measurement conditions as well as counting
statistics for each analysis. The laser was kept in one position for 100 s,
ablating a small pit (132 μm in diameter, approximately 50 to 100-μm
deep) in the bone. To address U-series age issues related to uncertainties
in open-system uranium uptake, the^230 Th/^238 U and^234 U/^238 U datasets
for each bone were fitted using a diffusion–absorption–decay model^68.
Coupled US–ESR dating of fossil teeth. Bovid molars were collected
for US–ESR direct dating, and sectioned using a large diamond-blade
rotating saw and polished to a 100-μm surface smoothness. Samples
NDG-1038, NDG-1163, NDG-2074, NDG-2562, NDG-2566 and NDG-2569
were first analysed by laser-ablation ICP-MS (LA-ICPMS) quadrupole
for uranium distribution, to assess the suitability of the samples for
US–ESR dating. Only the samples NDG-1038, NDG-1163 and NDG-2569
were found to be suitable, and these samples were prepared following
a previously developed protocol^66 (Extended Data Fig. 10). Each frag-
ment was then measured at Southern Cross University on a Freiberg
MS5000 ESR X-band spectrometer, and irradiated with the Freiberg
X-ray irradiation chamber. ESR intensities were extracted from the
merged spectra obtained from angular variation measurements^69 ,^70
(for an example, see Extended Data Fig. 10b), after correcting for
baseline, subtraction of isotropic signals and assessment of the con-
tribution of non-oriented CO 2 − radicals (NOCORs) using previously
published protocols^66 ,^70 (for an example, see Extended Data Fig. 10c).
Dose–response curves were obtained using the MCDOSE 2.0 soft-
ware^71 (Extended Data Fig. 10a). U-series dating was conducted on
both dentine and enamel at the University of Wollongong, using an
ESI NW193 ArF Excimer laser coupled to a MC-ICP-MS Neptune Plus to
calculate the internal dose rate. All age calculations were carried out
with the US–ESR program^72 , which uses previously published^73 dose
rate conversion factors.
Modelling of landscape, terrace and excavation chronologies
To evaluate the uncertainties of the integrated dating approach of
the landscape, terrace and fossil contexts (Supplementary Tables 6,
9, 10, 12, 13), Bayesian modelling was performed on all independent
age estimates using the OxCal (version 4.2) software^74 available at
https://c14.arch.ox.ac.uk/oxcal.html (Supplementary Tables 14, 15).
The analysis incorporated the probability distributions of individual
ages, and constraints imposed by stratigraphic relationships and the
reported minimum or maximum nature of some of the individual age
estimates. Each individual age was included as a Gaussian distribution
(with mean and s.d. defined by the age estimate and their associated
uncertainties). The U-series profiling ages on the fossil bone yielded a
range of ages (160–60 kyr) and these were incorporated as a uniform
distribution over this interval.
Data reporting
No statistical methods were used to predetermine sample size. The
experiments were not randomized and investigators were not blinded
to allocation during experiments and outcome assessment.
Reporting summary
Further information on research design is available in the Nature
Research Reporting Summary linked to this paper.